For long-life systems in aerospace, rail, medical and demanding industrial markets, the speed of semiconductor change collides with products designed to last decades. The result is a chronic engineering challenge. The good news is there is a way to design through it and services that can carry proven platforms forward without forcing hardware changes or software rewrites. This is the ultimate sustainability win: less resource, less waste, longer life.
Rochester Electronics has made a business out of keeping systems shipping and working. Its model spans authorised inventory, assembly, test and (when there is no part or die left) authorised recreation of the original silicon for a drop-in replacement. The approach is pragmatic. It starts with what exists, then escalates as necessary, always trying to avoid software impact for the OEM.
At the sharp end of this work is a set of choices engineers control on day one: which components sit closest to system software; which packages and assembly processes they rely on; and how they archive design data for future recovery. Those choices define the cost of later changes; maintainability of the bill-of-materials; and feasibility of sustaining lifetime performance.
Rochester Electronics’ director of sales for Northern Europe, Middle East and Africa, Paul Green, said: “Customers today are having a similar issue to what they had 20 years ago, in terms of how do I keep my products alive.”
Rochester Electronics’ VP of design and technology, Dan Deisz, added: “We are the method of last resort within Rochester and that’s the creation of new silicon that looks like the original silicon in a fully authorised way.”
Why obsolescence keeps accelerating for long-life systems
Green reported that obsolescence challenges are intensifying as device roadmaps compress and packaging, wafer nodes and tester platforms evolve. Many customers still support very old programmes while wishing to avoid redesign. The gap between long service life and short component lifecycles is growing, which raises the urgency for strategies that minimise redesign and preserve reliability.
Green was blunt about the trend: “Obsolescence is always going to be with us because technology always looks to move on. That is why Rochester positions itself to offer inventory where possible; then assembly and test from stored die and IP; and finally, silicon recreation when nothing else remains.”
Deisz framed the four root causes of obsolescence engineers should track: the fab process disappears, the package type disappears, the tester platform disappears, or the part fails to meet the manufacturers revenue targets. Those root causes determine next steps. Misjudged last-time buys, underestimated demand tails and new regulatory or programme requirements often push systems into a bind.
From a technology perspective, today’s landscape is very different to the component types found in legacy hardware. Deisz contrasted the legacy world with today’s proliferation of processor types, GPUs and microcontrollers. That diversity expands choice but increases variation in sustainment.
How an authorised physical clone avoids software change
When no finished goods or wafer die exist, Rochester’s last-resort method is a physical clone of the original die, executed with the authorisation of the original manufacturer. The team starts with the GDSII physical design database rather than HDL sources.
Deisz said: “We’re physically cloning the die, moving it from one fab process to another, duplicating the die size, duplicating the transistor structure, duplicating the interconnect and then manipulating it in place to perform the same. From the customer’s view the edge rates, power, and package are matched so the part drops in with no software changes.”
That clone is only one stage in a chain. Package availability is its own obsolescence vector. Test programme continuity is another. Where an original test programme exists, Rochester manages it. Where it does not, the team recreates what is needed alongside the physical design work, particularly for mature 5.0, 3.3 and 1.8 V designs where full-chip SPICE analysis is feasible.
As Deisz explained, GDSII sits at the heart of this method: “GDSII has been the ubiquitous physical design format for all semiconductor chip designs for decades. That persistence is exactly why it is suitable for long-term archiving strategies engineers can rely on when they need to sustain a design in the future.
Packaging choices that improve sustainment
Package choice is a sustainability decision when the aim is lifetime extension. Deisz highlighted that lead-frame packages that require trim and form are disappearing: “Everything should be BGA and QFN, if possible, because those assembly flows remain viable where many curved-lead formats will fade. This is not a subtle optimisation. If a design depends on a package style that assembly houses step away from, the risk to uptime and maintainability is immediate.”
Rochester invests selectively to keep certain package types alive but no single company can preserve all legacy styles. The practical message for design teams is to favour package types with strong forward manufacturing support. That decision interacts with the rest of the sustainment stack: tester availability, assembly capacity and the feasibility of reproducing die performance without perturbing software.
Archives that actually work, 20 years later
Design for sustainability includes design for recovery. Deisz argued that design archives often fails because they capture proprietary tool states that are impossible to recreate decades later. Teams mirror directories, then discover they need the exact operating system and toolchain versions from long-retired environments. He recommended tool-independent archiving: keep GDSII for the chip, use non-proprietary formats such as text/common image formats and store data on media that remains accessible rather than carriers with ambiguous life.
Deisz was candid about the status quo. When asked what percentage of design teams archive to that standard, he answered ‘next to zero’, including teams within semiconductor companies. He noted no ISO standard exists for semiconductor chip design archiving and that long-term system teams are rarely given the time or resources to achieve it. That gap directly undermines reuse, upgradeability and repairability in the field, because future engineers lack the artefacts needed to sustain the platform efficiently.
The sustainability implications are practical rather than rhetorical. Good archives improve maintainability and reduce rework. They also shorten the path to authorised recreation if all else fails. That persistence supports reliability over the service life and minimises the risk of wide-scale redesign that can cascade into fresh materials, new tooling and higher lifecycle emissions. The sustainability lexicon calls this lifetime extension a form of reuse and upgradeability that avoids unnecessary rebuilds.
Keep software out of the blast radius
In long-term systems, hardware changes are costly. Software changes are worse. Deisz pointed to studies and industry practice that show software changes as an order of magnitude more expensive than hardware changes for long-life programmes. The crucial takeaway for engineers is to identify the components that sit closest to system software—processors, ASICs, specialised RF etc—and prioritise those on the critical parts list. If those go end of life, a system can be stranded.
Green added that many customers arrive late, sometimes after outsourcing design or production and losing control over key details. Rochester’s staged model helps: first check inventory. Then manufacture from stored die and IP at low MOQs suitable for high-reliability markets. Only then escalate to recreation. The earlier a team engages, the more choices remain.
Green urged: “Let’s do a little bit of planning on what are those real critical parts because late engagement narrows options, raises cost and extends timelines.”
As a concrete example of where Rochester invests to protect software, Deisz described major investment in NXP PowerPC processors used by long-term system companies. Wafer fabrication continues at NXP, die is stored at Rochester, packages are controlled and tests run on the original platforms, all with the intention of keeping the portfolio running for decades. This is a template: preserve execution where the software would otherwise be at risk.
Sovereignty, trust, and what to watch in the market
For defence, security, and safety-critical sectors, sovereignty over the supply chain is essential. Green stressed the need for robust, traceable, reliable flows. Deisz extended the point: sometimes the only way to ensure trust is to bring design, assembly and test inside a single controlled organisation. That internalisation is not cosmetic, it is how guardrails are enforced when the cost of failure is unacceptable.
The company also cautioned against over-reliance on generic bill-of-materials health tools. These services can miss turning points that insiders see first, such as assembly house exits from legacy packages or impending end-of-life decisions. Direct dialogue fills that gap. Rochester’s teams work between suppliers and customers to spot end-of-life winds early, propose prior iterations where suitable and shape investments in design, assembly and test to keep critical devices available.
One useful market signal is automotive. Deisz said it is the only long-term market that reliably drives new product introductions at semiconductor companies. For long-life systems even in non-automotive sectors, aligning with vendors deeply committed to automotive can reduce long-term availability risk. He contrasted vendors that sprint in fast-moving segments with those oriented to long service life, advising teams to choose partners accordingly and to give design teams enough time to act on that requirement.
A practical to-do list for engineers starting today
- Start with a critical parts list oriented around software proximity. Identify the processor, ASIC and any specialised devices that would force a software event if changed. Treat those as the highest sustainment priority. Deisz said: “The closer a part is to system software, the more critical it is because touching software is the superset challenge to be avoided.”
- Choose packages for longevity, not novelty. Where possible, prefer BGA and QFN over legacy lead-frame types that require trim and form, which are disappearing. This is a maintainability decision that reduces the risk of assembly discontinuities during the service life.
- Design an archive that future engineers can actually open. Keep GDSII for the chip. Use non-proprietary formats for documents and images. Store data on media with known preservation characteristics. Plan for people who are not the original team to pick it up twenty years from now.
- Schedule regular market check-ins. Green suggested quarterly engagement as a baseline in a fast-changing landscape where last-time-buy windows can shrink from years to months. Early conversations keep options open, allow inventory to solve problems cheaply, and surface sustainment paths before a crisis.
- Finally, build sustainment into requirements and give the team time. Deisz urged leaders to write the long-term goal into the design brief, then allow the time needed to align suppliers, packages and archives to that aim. That investment strengthens reliability, reduces forced redesign and supports efficient reuse and upgradeability across the service life.
If you are already facing a critical parts issue, start with Rochester’s staged path. Confirm whether authorised inventory can meet the need. If not, explore assembly and test from stored die, then evaluate authorised recreation for a drop-in replacement that avoids software change. Engage early, share the artefacts you do have and treat the archive as an enabler rather than an afterthought. The sustainability of long-life systems depends on decisions made now.